InferMesh logo
Changelog

Changelog

Track feature launches, performance improvements, and bug fixes in every release.

v2.8.02026-03-14

Gemini 2.5 Pro Added, Audio Input Supported

  • ModelAdded Google Gemini 2.5 Pro (1M context, multi-modal)
  • ModelAdded Gemini 2.5 Flash lightweight model
  • NewAudio input support (audio modality), send base64 audio chunks directly
  • ImprovedStreaming latency optimized, P99 reduced from 80ms to 48ms
  • FixFixed parsing errors when function calling returned empty arguments
v2.7.22026-03-01

DeepSeek-R1 Integration & Stronger Fallback Strategy

  • ModelAdded DeepSeek-R1 reasoning model (128K context)
  • NewAdded cost-optimized fallback mode to prioritize the lowest price model with matching capability
  • ImprovedRefactored API key permission system to support model-group authorization
  • FixFixed inaccurate usage.total_tokens calculation in some cases
  • FixFixed occasional timeout on `/v1/models`
v2.7.02026-02-15

Claude 3.7 Sonnet Launched & Console Refresh

  • ModelAdded Anthropic Claude 3.7 Sonnet (200K context)
  • ModelAdded Claude 3.7 Haiku fast model
  • NewNew console experience with usage charts and model-level aggregation
  • NewTop-up records can now be exported as CSV
  • ImprovedImproved load-balancing with latency-aware distribution across providers
  • BetaTeam accounts in beta with member management and shared balance
v2.6.52026-02-01

GPT-5.1 & o3-mini Support

  • ModelAdded OpenAI GPT-5.1 (new flagship, 1M context)
  • ModelAdded OpenAI o3-mini reasoning model
  • ImprovedVision input now supports both URL and base64 formats
  • FixFixed missing final-frame finish_reason in streaming mode
v2.6.02026-01-18

Custom Rate Limits & Multi-region Deployment

  • NewPer-key RPM/TPM limits can now be configured in console
  • NewAdded Asia (Singapore) ingress node for lower latency
  • ImprovedError responses now follow standard OpenAI format with x-request-id
  • ModelAdded Mistral Large 2501 & Mistral Small 3
  • FixFixed extra escape characters in Claude JSON mode responses
v2.5.02026-01-02

Official Platform Launch

  • NewOfficial launch with first batch of 120+ models
  • NewSupports five providers: OpenAI, Anthropic, Google, DeepSeek, and Meta
  • NewProvides Chat Completions, Embeddings, and Models APIs
  • NewSupports Function Calling, JSON Mode, Vision, and Streaming
  • NewCore console features: API key management, balance view, and usage stats

Want update alerts first?

Create an account and we will notify you by email when major releases go live.

Sign Up FreeView Docs